Propagate failures in pandas integration tests and Skip failing tests

.github/workflows/pr.yaml

@@ -40,6 +40,7 @@ jobs:
       - pandas-tests
       - pandas-tests-diff
       - telemetry-setup
+      - third-party-integration-tests-cudf-pandas
     secrets: inherit
     uses: rapidsai/shared-workflows/.github/workflows/[email protected]
     if: always()
@@ -325,6 +326,17 @@ jobs:
         node_type: cpu4
         build_type: pull-request
         run_script: "ci/cudf_pandas_scripts/pandas-tests/diff.sh"
+  third-party-integration-tests-cudf-pandas:
+    needs: wheel-build-cudf
+    secrets: inherit
+    uses: rapidsai/shared-workflows/.github/workflows/[email protected]
+    with:
+      build_type: pull-request
+      node_type: "gpu-v100-latest-1"
+      arch: "amd64"
+      container_image: "rapidsai/ci-conda:latest"
+      run_script: |
+        ci/cudf_pandas_scripts/third-party-integration/test.sh python/cudf/cudf_pandas_tests/third_party_integration_tests/dependencies.yaml
 
   telemetry-summarize:
     runs-on: ubuntu-latest

ci/cudf_pandas_scripts/third-party-integration/run-library-tests.sh

@@ -23,8 +23,6 @@ runtest() {
     -v \
     --continue-on-collection-errors \
     --cache-clear \
-    --numprocesses=${NUM_PROCESSES} \
-    --dist=worksteal \
     ${TEST_DIR}/test_${lib}*.py
 }
 
@@ -42,8 +40,6 @@ main() {
     -v \
     --continue-on-collection-errors \
     --cache-clear \
-    --numprocesses=${NUM_PROCESSES} \
-    --dist=worksteal \
     ${TEST_DIR}/test_${lib}*.py
 }
 

ci/cudf_pandas_scripts/third-party-integration/test.sh

@@ -26,6 +26,8 @@ main() {
     LIBS=${LIBS#[}
     LIBS=${LIBS%]}
 
+    ANY_FAILURES=0
+
     for lib in ${LIBS//,/ }; do
         lib=$(echo "$lib" | tr -d '""')
         echo "Running tests for library $lib"
@@ -56,10 +58,6 @@ main() {
         rapids-logger "Check GPU usage"
         nvidia-smi
 
-        EXITCODE=0
-        trap "EXITCODE=1" ERR
-        set +e
-
         rapids-logger "pytest ${lib}"
 
         NUM_PROCESSES=8
@@ -72,12 +70,20 @@ main() {
             fi
         done
 
+        EXITCODE=0
+        trap "EXITCODE=1" ERR
+        set +e
+
         TEST_DIR=${TEST_DIR} NUM_PROCESSES=${NUM_PROCESSES} ci/cudf_pandas_scripts/third-party-integration/run-library-tests.sh ${lib}
 
+        set -e
         rapids-logger "Test script exiting with value: ${EXITCODE}"
+        if [[ ${EXITCODE} != 0 ]]; then
+            ANY_FAILURES=1
+        fi
     done
 
-    exit ${EXITCODE}
+    exit ${ANY_FAILURES}
 }
 
 main "$@"

python/cudf/cudf_pandas_tests/third_party_integration_tests/tests/test_stumpy_distributed.py

@@ -41,7 +41,7 @@ def test_multidimensional_distributed_timeseries(dask_client):
     rng = np.random.default_rng(seed=42)
     # Each row represents data from a different dimension while each column represents
     # data from the same dimension
-    your_time_series = rng.random(3, 1000)
+    your_time_series = rng.random((3, 1000))
     # Approximately, how many data points might be found in a pattern
     window_size = 50
 

python/cudf/cudf_pandas_tests/third_party_integration_tests/tests/test_tensorflow.py

@@ -271,6 +271,7 @@ def call(self, values):
         return tf.concat(values, axis=-1)
 
 
+@pytest.mark.xfail(reason="ValueError: Invalid dtype: object")
 def test_full_example_train_with_df(df, target):
     # https://www.tensorflow.org/tutorials/load_data/pandas_dataframe#full_example
     # Inputs are directly passed as dictionary of series

python/cudf/cudf_pandas_tests/third_party_integration_tests/tests/test_xgboost.py

@@ -123,7 +123,7 @@ def test_predict(device: str) -> np.ndarray:
 
     predt0 = reg.predict(X_df)
 
-    predt1 = booster.inplace_predict(X_df)
+    predt1 = booster.inplace_predict(X_df).get()
     np.testing.assert_allclose(predt0, predt1)
 
     predt2 = booster.predict(xgb.DMatrix(X_df))